Diffusion-Weighted Imaging (DWI) and Diffusion Tensor Imaging (DTI)

Why diffusion imaging matters

Brain function depends critically on connectivity, not just isolated cortical regions. White-matter tracts enable communication between distributed brain areas, and disruption of these pathways underlies many neurological and neuropsychological conditions.

Diffusion MRI tracks how water molecules move inside the brain. In white matter, water tends to move along nerve fibers, not across them. By measuring this movement, scientists can infer the direction, integrity, and organization of white-matter pathways, basically mapping the brain’s wiring while a person is alive.

Regular (structural) MRI, by contrast, mostly shows anatomy: the shape, size, and location of brain structures. It can tell you where white matter is, but not how well it’s connected or which regions are talking to each other.

What diffusion is

Diffusion refers to the random movement of molecules due to thermal energy. Molecules naturally move from areas of higher concentration to lower concentration.

In the brain:

• Higher diffusivity means water molecules move more freely

• Lower diffusivity means movement is restricted by tissue structure

Diffusion MRI measures how freely water molecules move within brain tissue.

Isotropic vs anisotropic diffusion

Isotropic diffusion

Diffusion is equal in all directions.
This occurs in:

• cerebrospinal fluid (CSF),

• grey matter (to a lesser extent).

There is no preferred direction of movement.

Anisotropic diffusion

Diffusion does not occur equally in all directions.
In white matter, the movement of water is directionally constrained.

This happens because:

• Axons are organized into parallel bundles

• Myelin sheaths and internal cellular structures limit water movement across fibers

• Water moves most freely along the length of axons and is hindered in directions perpendicular to them.

This directional dependence of diffusion is known as anisotropy, and it is the fundamental property that diffusion MRI measures and uses to characterize white-matter structure.

Diffusion properties in brain tissue

Different tissues show distinct diffusion characteristics:

• CSF shows very high diffusivity and very low anisotropy

• Grey matter shows moderate diffusivity and low anisotropy

• White matter shows moderate diffusivity but high anisotropy

These differences allow diffusion MRI to distinguish tissue types and infer white-matter organisation.

Diffusion-Weighted Imaging (DWI)

DWI is the basic diffusion MRI acquisition. It measures how much the MRI signal is attenuated by diffusion in different directions.

Key concepts:

• Diffusion sensitivity can be applied in any direction

• Signal attenuation depends on how freely water moves in that direction

• Acquiring data in multiple gradient directions allows inference about tissue structure

What DTI is

Diffusion Tensor Imaging (DTI) is a type of MRI that uses water movement to learn about white matter in the brain.

It does not see nerve fibres directly.
It infers their organisation from how water moves.

What DTI measures

DTI measures how water diffuses inside brain tissue, focusing on:

• How much water moves (overall diffusion)

• Whether water moves more in one direction than others (anisotropy)

• The main direction of water movement inside each voxel

From this, we calculate common measures:

• FA (fractional anisotropy) – how directional the diffusion is

• MD (mean diffusivity) – how freely water moves overall

• Principal direction – the dominant orientation of diffusion

How DTI works (step by step)

1. The MRI applies diffusion sensitivity in many directions.

2. Water movement reduces the MRI signal differently depending on direction.

3. These differences are measured inside each voxel.

4. A tensor model (a mathematical shape) is fitted to summarise diffusion.

5. The tensor describes:

   • diffusion strength

   • diffusion direction

   • how directional it is

The main axis of the tensor is assumed to line up with fibre orientation.

What DTI is good at (strengths)

• Works in living humans

• Non-invasive and safe

• Good at showing large, well-organised white-matter tracts

• Sensitive to:

  • development

  • injury

  • stroke

  • tumours

• FA and MD are easy to calculate and compare across people

• Forms the basis for tractography

DTI is especially useful for group comparisons and clinical applications.

What DTI cannot do (limitations)

• Voxel size is large (mm), axons are tiny (µm)

• Measures are indirect and reflect many tissue features at once

• Sensitive to damage, but not specific to the cause

• Cannot show direction of information flow

• Fails when fibres cross or fan within a voxel

• Assumes only one main fibre direction per voxel

• Tractography based on DTI can produce:

  • false connections

  • missing real connections

What MRI is (baseline)

MRI is a way of making images using:

• a strong magnetic field

• radio waves

• signals from hydrogen atoms (mostly in water)

Standard MRI (like T1 or T2 scans) shows:

• anatomy

• shapes and boundaries of tissues

It does not look at water motion.

What DWI is, and why it’s still MRI

DWI (Diffusion-Weighted Imaging) is MRI with an extra trick added.

How:

• the scanner applies special magnetic gradients

• these gradients make the signal sensitive to water movement

What changes:

• if water moves a lot → signal drops

• if water is restricted → signal stays stronger

Why it’s MRI:

• same scanner

• same physics

• just different gradient settings

So DWI is:

What DTI is, and how it relates to DWI

DTI (Diffusion Tensor Imaging) is not a new scan.

It is:

How:

• DWI images are acquired in many directions

• a mathematical model (the tensor) is fitted to those measurements

What DTI gives:

• FA (directionality)

• MD (overall diffusion)

• main diffusion direction

So:

• DWI = how the data are acquired

• DTI = how the data are modelled and summarised

Why DWI and DTI exist at all

Standard MRI shows:

• what tissue looks like

DWI/DTI show:

• how tissue is organised at a microscopic level

• especially white matter structure

They are used because:

• they detect stroke earlier than conventional MRI

• they are sensitive to white matter damage

• they allow tractography

One clean way to remember it

• MRI = the general imaging method

• DWI = MRI made sensitive to water movement

• DTI = a model applied to DWI to describe direction and structure

The key idea

Normal (structural) MRI sees shape and structure.
DWI sees microscopic water movement.

Some problems change water behaviour first, long before they change visible structure.

Why a normal MRI can miss early stroke

What happens in acute stroke (minutes–hours)

• Blood flow stops

• Cells lose energy

• Ion pumps fail

• Cells swell (cytotoxic oedema)

• Water gets trapped inside cells

At this point:

• Brain tissue has not fallen apart

• No bleeding

• No big swelling you can see

So on normal MRI:

• Shape looks normal

• Tissue boundaries look normal

• Nothing obvious to “see”

But at the microscopic level:

• Water movement has changed dramatically

DWI is sensitive to that immediately.

Why normal MRI sees tumours but not early damage

Tumours change structure

Tumours usually cause:

• extra tissue

• mass effect

• displacement of normal anatomy

• blood–brain barrier leakage

These are macroscopic changes.

Structural MRI is very good at:

• seeing abnormal shapes

• detecting contrast enhancement

• spotting mass effects

That’s why tumours are often obvious.

Why cellular swelling is invisible to normal MRI

Cellular swelling means:

• cells get a bit bigger

• extracellular space gets smaller

But:

• this happens at the micrometre scale

• MRI voxels are millimetres wide

So the voxel still looks “full of brain” either way.

Normal MRI averages everything together and sees:

DWI sees:

Why DWI sees things before anatomy changes

Normal MRI relies on:

• tissue breakdown

• fluid shifts

• bleeding

• structural distortion

Those take hours to days.

DWI relies on:

• water motion inside cells

That changes in minutes.

So:

Tumour characteristics and treatment response

Tumours are a special case because both methods help.

• Structural MRI:

  • shows size and location

• DWI:

  • shows how tightly packed the cells are

  • shows whether treatment is killing cells

A tumour can:

• look the same size

• but have very different diffusion properties

So DWI gives extra information, not redundant information.

One sentence to remember

Normal MRI and DWI look at different kinds of changes in the brain. A normal MRI shows structure and shape, so it is good at seeing things like tumours, bleeding, or major swelling, but it only changes when the tissue itself looks different. In conditions like an early stroke, the brain tissue still looks normal, but the cells stop working properly and swell, which traps water inside them. Normal MRI cannot see this because the changes are too small. DWI, however, is sensitive to how water moves, so it can detect these microscopic changes within minutes, long before the brain’s structure changes. That’s why doctors often use both scans together: one shows what the brain looks like, and the other shows how healthy the tissue is at a very small scale.

b-value

The b-value tells us how strongly the MRI is made sensitive to water movement.

• Low b-value → weak sensitivity to diffusion

• High b-value → strong sensitivity to diffusion

When the b-value is higher:

• differences in water movement are easier to see

• but the image becomes noisier (less signal)

That’s the trade-off:
better contrast, but poorer signal quality.

How b-values are used in practice

A diffusion scan usually includes:

• b = 0 images (no diffusion weighting, like a normal MRI)

• one or more higher b-values (to measure diffusion)

The scanner compares these images to work out diffusion.

Scan times are relatively short (minutes), making DWI practical in clinical settings.

Clinical usefulness of DWI

DWI is highly sensitive to acute tissue changes.

Key applications:

• Acute ischaemic stroke: restricted diffusion appears within minutes to hours, earlier than conventional MRI changes

• Tumours and oedema: diffusivity is altered by changes in cellular density and extracellular space

• Monitoring treatment response: changes in diffusivity can reflect tumour progression or regression

Because diffusion responds quickly to microstructural change, DWI is a powerful clinical tool.

Artefacts and preprocessing in DWI

DWI uses fast imaging techniques, which makes it vulnerable to artefacts.

Common issues include:

• Susceptibility-induced distortion at tissue–air boundaries

• Eddy currents from rapidly switching gradients

• Patient motion, particularly relevant in paediatric imaging

Standard preprocessing tools (e.g. TOPUP and EDDY) are used to correct these distortions and recover anatomically meaningful data.

From DWI to DTI

DWI provides diffusion-weighted images.
DTI (Diffusion Tensor Imaging) adds a mathematical model to describe diffusion within each voxel.

The diffusion tensor models diffusion as an ellipsoid, characterised by:

• the amount of diffusion,

• the degree of anisotropy,

• the principal direction of diffusion.

DTI therefore summarises diffusion information rather than just displaying raw signal attenuation.

DTI summary parameters

Fractional anisotropy (FA)

FA quantifies how directional diffusion is.

• FA ≈ 0 → isotropic diffusion

• FA closer to 1 → highly anisotropic diffusion

High FA is typically associated with organised white-matter tracts.

FA (Fractional Anisotropy) is important because it tells us how organised and directional white-matter tissue is.

In white matter, axons are bundled together and wrapped in myelin. This structure forces water to move mainly along the fibres rather than in all directions. FA measures how strongly water movement is constrained to one direction. When white matter is healthy and well organised, FA is high.

When something disrupts that organisation, FA changes. If axons are damaged or lost, if myelin breaks down, if there is swelling, or if fibres cross within the same voxel, water movement becomes less directional and FA decreases. Because of this, FA is very sensitive to changes in white-matter microstructure, even when the brain still looks normal on standard MRI.

Mean diffusivity (MD)

MD reflects the overall magnitude of diffusion, regardless of direction.

• Increased MD can reflect oedema, tissue loss, or inflammation

• Decreased MD can reflect acute cytotoxic oedema (e.g. stroke)

MD (Mean Diffusivity) is important because it tells you how freely water can move in brain tissue, which reflects tissue health at a microscopic level.

Water moves differently depending on what the tissue is like. When tissue is healthy and intact, cell membranes, axons, and myelin act as barriers, so water movement is partly restricted and MD stays in a normal range. When tissue is acutely injured, such as in an early stroke, cells swell and trap water, so water movement is reduced and MD decreases. When tissue is damaged or breaking down over time, barriers are lost, water moves more freely, and MD increases.

Principal diffusion direction (PDD)

The PDD represents the direction of greatest diffusion within a voxel.

By assuming this aligns with the dominant fibre orientation, we can estimate local white-matter direction.

PDD (Principal Diffusion Direction) is important because it tells us the main orientation of white-matter fibres in each small region of the brain.

In white matter, water moves most easily along the length of axons. When DTI measures diffusion in many directions, it finds the direction where water moves the most. That direction is the principal diffusion direction.

This matters because:

• it tells us the local orientation of white-matter bundles

• it is the building block of tractography (connecting directions across voxels)

• PDD assumes one main fibre direction per voxel, so it struggles where fibres cross or fan.

• PDD is important because it gives the main fibre direction in each voxel, enabling mapping and tracking of white-matter pathways.

FA colour maps

FA colour maps encode the orientation of white matter:

• Red = left–right

• Green = anterior–posterior

• Blue = superior–inferior

Importantly, these maps show orientation only, not direction of information flow.

Tractography

Tractography reconstructs white-matter pathways by:

1. selecting seed points,

2. following the principal diffusion direction voxel-to-voxel,

3. stopping when diffusion becomes isotropic or curvature is excessive.

This allows visualisation and segmentation of major white-matter tracts.

Applications include:

• white-matter mapping,

• studying effects of lesions or tumours,

• surgical planning and neuronavigation,

• tract-based group analyses of FA and MD.

Noise, uncertainty, and ROIs

Tractography accumulates local errors as it progresses, and trajectories can jump incorrectly between fibre bundles.

To manage this:

• probabilistic tractography estimates uncertainty,

• regions of interest (ROIs) are used to constrain tracking,

• waypoint, exclusion, and termination masks improve anatomical specificity.

ROI definitions are not easily transferable between individuals, but standardised protocols (“recipes”) can be applied.

Limitations of DWI and DTI

Diffusion MRI has important limitations:

• Spatial resolution is relatively low (2–3 mm voxels)

• Diffusion measures are sensitive but not specific

• Changes may reflect many processes (degeneration, inflammation, oedema)

• Direction of communication cannot be inferred

• Tractography works best for large, coherent tracts

• False positives and false negatives are common

The crossing-fibre problem

DTI assumes a single dominant fibre orientation per voxel.
In reality, many voxels contain crossing or kissing fibres.

When fibres cross:

• there is no true single principal direction,

• FA can be misleading,

• tractography may fail.

More advanced models (e.g. HARDI, multi-fibre models) attempt to address this, but introduce new trade-offs.

Integrated take-home messages

• Diffusion MRI exploits the directional movement of water to probe white-matter structure.

• DWI measures diffusion sensitivity; DTI adds a tensor model.

• FA, MD, and PDD summarise different aspects of diffusion.

• Diffusion MRI is clinically powerful but biologically non-specific.

• Tractography visualises connectivity but must be interpreted cautiously.

• Developmental stage critically shapes diffusion measures.

Final synthesis sentence

Diffusion MRI provides a uniquely sensitive window into white-matter organisation and development, but meaningful interpretation requires understanding the physics, modelling assumptions, developmental context, and limitations of DWI, DTI, and tractography.

Tractography: what it is

Tractography is a method for reconstructing white-matter pathways using diffusion MRI.

It works by:

• following the main diffusion direction from voxel to voxel

• linking these directions together to form lines called streamlines

When many streamlines run together, they form fibre bundles, which often correspond to known white-matter tracts.

This allows researchers and clinicians to:

• visualise white matter

• isolate specific tracts

• measure values like FA or MD within those tracts

Important point: tractography is a model-based reconstruction, not a direct image of axons.

Noise and uncertainty in tractography

Diffusion measurements are noisy, and each voxel has some uncertainty.

As tractography follows directions step by step:

• small local errors add up

• streamlines may slowly drift away from the true pathway

• paths can jump from one tract to a nearby one

Probabilistic tractography addresses this by:

• running tracking many times

• showing a range of possible paths

• giving an idea of uncertainty rather than one “correct” line

Regions of interest (ROIs)

ROIs are used to constrain tractography and make results more anatomically meaningful.

They can:

• define where streamlines start (seed)

• force streamlines to pass through certain areas (waypoints)

• stop or exclude streamlines from unwanted regions

Using ROIs:

• reduces false connections

• is standard practice in applications like surgical planning

Complex ROI strategies

Often, one ROI is not enough.

Multiple ROI types may be combined:

• seed regions

• waypoint regions

• exclusion masks

• termination regions

Challenges:

• ROIs are hard to transfer directly between people

• brains differ in size, shape, and development

However:

• standardised ROI “recipes” can be reused across studies

Development and ageing: why diffusion measures change

White matter does not stay the same across life — it develops, matures, and later declines. Diffusion MRI is sensitive to these changes.

During childhood and adolescence

The brain is still building and refining its white matter.

What happens biologically:

• Myelination increases (axons get more insulation)

• Fibres become more tightly packed and better aligned

• Connections become more efficient

What diffusion MRI sees:

• FA increases
  Water is pushed more strongly along fibres as organisation improves.

• MD decreases
  Extra barriers from myelin and cells restrict water movement.

Many white-matter tracts mature at the same time, so their diffusion measures often change together.

During ageing

With ageing, some of these processes reverse.

What happens biologically:

• Myelin breaks down

• Axons may be lost or become disorganised

• Tissue becomes less dense

What diffusion MRI sees:

• FA decreases
  Water movement becomes less directional.

• MD increases
  Water moves more freely as barriers are lost.

Limitations of diffusion MRI and tractography

Key limitations include:

• relatively low spatial resolution (millimetres)

• diffusion measures reflect many tissue properties at once

• sensitive to damage, but not specific to the cause

  • degeneration, inflammation, oedema can look similar

• cannot show direction of information flow

• tractography works best in large, well-organised tracts

• false positives and false negatives are common

Limitations of DTI specifically

DTI uses a single tensor per voxel.

This works well when fibres:

• are parallel

• dominate the voxel

But it fails when:

• two or more fibre bundles cross within one voxel

• there is no single true principal direction

In these cases, DTI oversimplifies the anatomy.

The crossing-fibre problem

In real brains:

• multiple fibre directions in one voxel are very common

Because the tensor model can only represent one direction:

• crossing fibres are misrepresented

• some tracts are missed or incorrectly reconstructed

This is a fundamental limitation of DTI, not a scanning mistake.

Improved models beyond DTI

More advanced diffusion models:

• use more diffusion directions

• often use higher b-values

• allow multiple fibre orientations per voxel

Advantages:

• better handling of crossings

• more realistic fibre orientation estimates

Disadvantages:

• longer scan times

• lower signal

• more false-positive connections

• diminishing returns beyond a point

Big-picture summary

• Water diffusion in the brain is shaped by tissue structure.

• Diffusion MRI measures this indirectly.

• Tractography follows preferred diffusion directions to estimate white-matter pathways.

• Diffusion measures are useful for studying development, stroke, tumours, and surgical planning.

• All results require careful interpretation, because models simplify complex biology.

Fractional Anisotropy (FA): what changes mean

One line to remember

White matter and water movement (very simple)

White matter is made of long nerve fibres (axons) wrapped in myelin.
These fibres run together in bundles, like straws in a box.

Because of this structure:

• water can move more easily along the fibres

• water has difficulty moving across the fibres

So water movement is directionally constrained in white matter.

What “water diffusion step size” means

Water molecules are always moving randomly.

The water diffusion step size means:

In the brain, this distance is:

• only a few micrometres (µm)

• roughly the size of a cell or smaller

At this tiny scale, water molecules:

• constantly bump into axons and myelin

• are guided by how the fibres are arranged

What a structural MRI voxel is

MRI does not look at single cells.

Instead, it divides the brain into many tiny 3D boxes called voxels.

A structural MRI voxel is:

Each voxel contains:

• thousands of axons

• lots of water

• different types of tissue all mixed together

MRI measures the average signal from everything inside one voxel.

Why size matters

This is the key idea.

• Water moves over a few micrometres

• MRI measures signals from millimetre-sized voxels

So:

• MRI cannot see individual axons

• MRI cannot see the tiny paths water takes

But the overall pattern of water movement inside a voxel is affected by those tiny structures.

What DWI actually measures

Diffusion-weighted imaging (DWI) measures:

Even though the fibres are too small to see:

• their presence changes how water moves

• this change shows up in the diffusion signal

So:

That means:

• not directly

• but as an average effect across many molecules and fibres

One analogy to lock it in

Imagine shaking a sealed box:

• inside are thousands of long sticks

• you can’t see them

• but the box resists motion more in some directions

From that resistance, you infer: “There must be aligned objects inside.”

DWI does the same thing using water motion.

One sentence to remember

Water moves over micrometre distances, MRI measures millimetre-sized boxes, and diffusion MRI uses the average water motion to infer structures that are too small to see directly.

Pros and Cons of MRI, DWI, DTI, and Tractography